[gmx-users] Protein angles / single point calculations

[Mark Abraham]

On 12/01/2012 9:14 PM, Matthias Ernst wrote:
> Hi,
>
> I have some questions.
> First, what dihedral angles will Gromacs produce and use in topology 
> when simulating proteins? To be more specific, I was looking for psi 
> (N-C(a)-C-N) and phi (C-N-C(a)-C) angles in all the dihedrals that 
> were generated by pdb2gmx. I only found the psi angles but not phi. Is 
> that by purpose?

If your force field doesn't implement such dihedrals, then they won't be 
there. However it seems to me more likely that you aren't looking in 
enough right places. Consider the opposite order of the atom numbers, also.

> Second, for testing purposes, I want to do single point energy 
> calculations in vacuo with no cutoffs at given coordinates. When I use 
> an mdp like the one below, it always changes the coordinates slightly 
> but enough to change the energy values noticeably. Is there a better 
> way to to this?

Zero-step MD is a way to do this. Using constraints = none, or 
continuation = yes will stop slight changes from the use of constraints. 
Another alternative is using mdrun -rerun, which is efficient for doing 
many such evaluations, and guarantees no changes to the coordinates.

> Third, I'm not sure if the coordinates are used correctly. To compare 
> the energy values, I inserted a little piece of code in the energy 
> evaluation routines which outputs the coordinates of all atoms. It 
> seems that everything beyond the three digits after the decimal sign 
> in the .gro files is filled with arbitrary numbers. I know that using 
> floating point data types results in a loss of accuracy but I thought 
> even with single precision, the accuracy should be higher than three 
> digits?

The accuracy is as high as it can be - I think you're assuming nothing 
has happened, and that isn't quite true.

Mark

>
> Here is my mdp:
> constraints   =       all-bonds
> integrator    =       MD
> nsteps          =       0
> nstcomm     =       1
> ns_type        =       grid
> rlist              =       0.0
> rcoulomb     =       0.0
> rvdw            =       0.0
> Tcoupl          =       no
> Pcoupl         =       no
> gen_vel        =       no
> pbc              =       no
>
> I would greatly appreciate any help and comments.
>
> Best regards,
> Matthias Ernst